Prompt
Everything you send to an AI model: the question, the instructions, the context.
A prompt is everything you send to an LLM before it answers: your question, any instructions ('answer in three bullet points'), examples of what you want, and material to work with (an email to summarize, a table to read). The model has no memory of you beyond what is in the prompt, so for that request the prompt is its whole world.
Small changes in the prompt change the answer a lot. 'Write a summary' and 'Write a two-sentence summary for a busy doctor, no jargon' produce very different text. Saying who it is for, what format you want, how long, and what to avoid gets better results than hoping the model guesses. This craft is called prompt engineering, though most of it is just clear writing.
Products built on LLMs add a hidden part, the system prompt: instructions from the developer that sit before yours and set the tone, the rules and the persona. When a chatbot answers in a particular style or declines something, the system prompt is often the reason.
Why it matters: with the same model, the quality of what you get depends mostly on what you ask. Learning to write a good prompt is the cheapest way to get more out of any AI tool.
Keep in mind
- The prompt is all the model knows about your request: say what you want, for whom, in what format.
- Showing one or two examples often beats describing the format.
- System prompts are hidden instructions from the developer that shape every answer.
- Prompt and answer are both measured in tokens, which is what you pay for.
